If you're looking for a standalone workstealing queue class implementation in C++ built on pthread or boost::thread, good luck, to my knowledge there isn't one.
However, as others have said Cilk, TBB and Microsoft's PPL all have workstealing implementations under the hood.
The question is do you want to use a workstealing queue or implement one? If you just want to use one then the choices above are good starting points simply scheduling a 'task' in any of them will suffice.
As BlueRaja said the task_group & structured_task_group in PPL will do this, also note that these classes are available in the latest version of Intel's TBB as well. The parallel loops (parallel_for, parallel_for_each) are also implemented with workstealing.
If you must look at source rather than use an implementation, TBB is OpenSource and Microsoft ships the sources for its CRT, so you can go spelunking.
You can also look on Joe Duffy's blog for a C# implementation (but it's C# and the memory model is different).
-Rick
